Miles Redcliffe and Matthew Blackwood 2026. Improving Interpretability in Large Language Models through Layer Wise Relevance Propagation for Identifying Latent Reasoning Biases. International Journal of Artificial Intelligence Research. 1, 2 (May 2026).